Virginia’s artificial intelligence registry faces transparency challenges, study shows

This story was originally published by the Virginia Mercury.

Continuity gaps within Virginia’s artificial intelligence registry are raising concerns about transparency, oversight and the implementation of standards across state government, state researchers said Monday. 

JLARC provided lawmakers an update on the registry as part of an ongoing review of the Virginia Information Technologies Agency (VITA), which services 65 executive branch agencies, security, project management, procurement and information technology planning for the commonwealth. The registry’s purpose is to track and approve state projects that would use AI technology.  

Researchers said that as of Oct. 31, VITA approved 90 out of about 182 cases that were submitted by state agencies frustrated with the agency’s process guidance and the registry using multiple systems that are “not well coordinated.” VITA said the total approved registry AI cases increased to 112 at Monday’s meeting.

“One theme that emerged through all of these conversations is that VITA does not have one singular voice around AI for agencies,” said Kimberly Sarte, associate director for ongoing oversight and fiscal analysis for the Joint Legislative Audit and Review Commission (JLARC). “We heard the different teams within VITA may have different answers to the same question depending on when they were asked.” 

Sarte added that VITA is working to establish a “centralized AI information role, which will also bring them in line with our peer states.”

VITA was required to establish IT security standards under an executive order signed by Gov. Glenn Youngkin on Jan. 18, 2024.

From the order, VITA created Virginia’s AI registry for agencies to submit potential AI use cases, or projects, for review and approval by VITA. Cases are approved based on whether there is a “positive outcome for citizens,” according to the state standards published by VITA.

The state standards also require agencies to examine alternative technologies and processes before choosing AI, to develop a regulatory impact analysis, and to clearly state whether AI will make recommendations or decisions on behalf of users. 

Sarte said the AI tools Smartling and CoPilot Chat have been made available to all executive branches.

Some of the approved 90 cases include the Virginia Racing Commission using meeting transcription for hearings, the Department of Corrections improving facility monitoring and security and VITA developing a chatbot to help users navigate its website.

However, the problem researchers identified is that case status is not updated regularly, and some agencies are unclear about which types of use cases to include and are not reporting them.

Researchers are also unclear whether the cases not entered into the registry comply with state standards.

Sarte said VITA is working to address some of the frustrations including developing an AI resource team to reduce inconsistent guidance.

Bob Osman, chief information officer for VITA, did not directly address the frustrations expressed by agencies. However, he did say the registry is a key part of the agency’s approach to AI oversight.

On Monday, Osman presented a chart that indicated that the use of AI is accelerating. AI is currently being used to help the Department of Transportation determine when roads need to be paved, to aid the Department of Quality review and approve water permits faster and help VITA create programs faster with fewer mistakes, improving government technology tools more efficiently.

“If you look at the results, we’ve come a long way in terms of where we got started, and in terms of the number of use cases (that) are approved in Virginia,” Osman said, “and we’re finding more and more use cases as agencies look at opportunities and try to learn about ways of taking advantage and harnessing this really powerful technology.”

He added that “every job is going to involve some kind of AI. There’s just no way around it.”

Relatedly, Democratic U.S. Sen. Mark Warner, who represents Virginia, and Republican Sen. Josh Hawley from Missouri introduced bipartisan legislation called the AI-Related Jobs Clarity Act. The bill aims to track jobs lost to AI as part of a broader effort to increase transparency about AI use in government. 

Under the proposed legislation, major companies and federal agencies would be required to report AI-related layoffs to the U.S. Department of Labor, which would compile them into a publicly available report.

“Good policy starts with good data,”  Warner said in a Nov. 5 statement. “This bipartisan legislation will finally give us a clear picture of AI’s impact on the workforce – what jobs are being eliminated, which workers are being retrained, and where new opportunities are emerging. 

The goal is to “make sure AI drives opportunity instead of leaving workers behind,” Warner said.

Virginia Mercury is part of States Newsroom, a nonprofit news network supported by grants and a coalition of donors as a 501c(3) public charity. Virginia Mercury maintains editorial independence. Contact Editor Samantha Willis for questions: info@virginiamercury.com.